Wiring a Voice AI Agent into GHL Workflows
End-to-end snapshot: inbound voice → conversation AI → calendar booking → 2-way SMS follow-up.
The setup
Most agencies still treat Voice AI like a novelty — a toy that answers the phone. The 2026 release changed that. With the new `voice.agent` action node, you can hand the agent a full workflow context: pipeline stage, last touch, custom values, calendar slots.
The seven-node workflow
1. **Inbound call trigger** — set the inbound number to the agent. 2. **AI router** — classify intent: book, reschedule, support. 3. **Calendar lookup** — pull next 3 slots from a team calendar. 4. **Voice confirm** — agent reads the slots back, captures the pick. 5. **Booking action** — write to the calendar with the call recording as a note. 6. **2-way SMS** — send confirmation + add-to-calendar link. 7. **Slack notify** — drop the recording + transcript into #sales-live.
What surprised us
We assumed prospects would hate the agent. They didn't — they hated *bad* IVRs. Once the agent could actually book, our pickup-to-book rate jumped from 11% to 38% across three pilot accounts.
Pricing the snapshot
If you're selling this as a service, don't sell the snapshot — sell the booked appointments. Charge $40-$80 per booked call, cap the monthly retainer, and keep the snapshot proprietary.